Imprimitivity theorem for groupoid representations

IMPRIMITIVITY THEOREM FOR

wo:

< : GROUPOID REPRESENTATIONS

(N

Leszek Pysiak
Department of Mathematics and Information Science,

6.1

Warsaw University of Technology

> .

^ ■ Plac Politechniki 1, 00-661 Warsaw, Poland

O

(N

od
O
O

March 30, 2010

Abstract

We define and investigate the concept of the groupoid repre-
sentation induced by a representation of the isotropy subgroupoid.
Groupoids in question are locally compact transitive topological
groupoids. We formulate and prove the imprimitivity theorem for
such representations which is a generalization of the classical Mackey's
theorem known from the theory of group representations.

Keywords: Groupoids, Induced representations, Im-
primitivity systems.

Mathematical Subject Classifications: 22A22, 22A30,
22D30

1 Introduction

The present paper, devoted to the study of the theory of groupoid repre-
sentations, is a continuation of my previous work [20] in which one can find
a presentation of the groupoid concept and of the groupoid representation
concept, important examples as well as relationships between groupoid rep-
resentations and induced group representations (see also [18], [21], [23], [TT]).
Groupoids have now found a permanent place in manifold domains of math-
ematics, such as: algebra, differential geometry, in particular noncommuta-
tive geometry, and algebraic topology, and also in numerous applications,
notably in physics. It is a natural tool to deal with symmetries of more com-
plex natura than those described by groups (see [23], [11], [2]). Groupoid
representations were investigated by many authors and in many ways (see

m, Mi SB], CP. 0)-

In a series of works ([7], [9], [8], [19]) we have developed a model unify-

2

ing gravity theory with quantum mechanics in which it is a groupoid that
describes symmetries of the model, namely the transformation groupoid of
the pricipal bundle of Lorentz frames over the spacetime. To construct the
quantum sector of the model we have used a regular representation of a non-
commutative convolutive algebra on this groupoid in the bundle of Hilbert
spaces. In paper ([5J) we have applaied this groupoid representation to in-
vestigate spacetime singulariies, and in [10] the representation of the funda-
mental groupoid to the gravitational Aharonov-Bohm effect.

The present work is aimed at introducing the concept of the groupoid
representation induced by a representation of the isotropy subgroupoid. We
assume that the groupoid in question is a locally trivial topological groupoid
and as a topological space it is a locally compact Husdorff space (Section 4).
This concept, framed "in the spirit of Mackey" is a natural generalization
of induced represetation of locally compact groups, created and investigated
by him [15J. Representations, investigated in the present work, are unitary
and are realized in Hilbert bundles over the unit spaces of a given groupoid
|18j . Section 5 contains the formulation and the proof of the imprimitivity
theorem for groupoids which is a generalization of the classical Mackey's im-
primitvity theorem for group representations [13], [15]. The theorem says

3

that every unitary groupoid representation, for which there exists the im-
primitivity system, is a representation induced by a representation of the
isotropy subgroupoid.

In Section 6, I investigate induced representations of the transformation
gropoid over a homogeneous space of the group G and show that there exists a
strict connection between these representations and induced representations
of the group G (in the sense of Mackey.

In Section 7, I give a physical interpretation of concepts analyzed in Sec-
tion 6. I describthe a representation that has been used in the mentioned
above model unifying gravity and quanta when this model is reduced (as the
result of the act of measurement) to the usual quantum mechanics. And
then I consider the energy-momentum space for a massive particle (it is a
homogeneous space of the Lorentz group) and the transformation groupoid
corresponding to this space. I also give a definition (in the sense of Mackey
|14j . [13]) of a particle as an imprimitivity sestem for the unitary represen-
tation of this groupoid.

4

2 Preliminaries

Let Q be a groupoid over a set X (the base of £/). We recall (cf. [I], [18] )
that a groupoid ^ is a set with a partially defined multiplication "o" on a
subset Q 2 of Q x Q, and an inverse map g — > g^ 1 defined for every g G Q. The
multiplication is associative when defined. One has an embedding e : X — > Q
called the identity section and two structure maps d, r : Q — > X such that

for g E Q.

Let us introduce the following fibrations in the set Q:

G x = {g : d(g) = x}

G x = {geG: r(g) = x}

for x G X. Let us also denote Q y x = Q x f]Q y , and consider the set Q* =
Q x p| Q x for x G X. It has the group structure and is called the isotropy group
of the point x. It is clear that the set T = \J x£X Gx has the structure of a
subgroupoid of Q over the base X (all the structure maps are the restrictions
of the structure maps of Q to T).

5

We call Q a transitive groupoid, if for each pair of elements Xi,x 2 G X
there exists g G Q such that d(g) = x\ and r(g) = x 2 .

A groupoid Q is a topological groupoid if Q and X are topological spaces
and all structure maps are continuous (in particular, the embedding e is a
homeomorphism of X onto its image).

In the following we assume that Q (and thus X) is a locally compact
Hausdorff space.

Example 1 A pair groupoid. Let X be a locally compact Hausdorff
space. Take Q = X x X. We define the set Q 2 of composable elements
as Q 2 = {((x , y) , (y , z)) : x,y,z G X} C Q x Q and a multiplication, for
{(x,y),(y,z)) G Q 2 , by

(x,y) o (y, z ) = (x,z).

Moreover, we have: (x,y)^ = (y,x), d(x,y) = y, r(x,y) = x, e(x) = (x,x).
With such defined structure maps Q is a groupoid, called pair groupoid.

Example 2 A transformation groupoid. Let X be a locally compact Haus-
dorff space, and G a locally compact group. Let G act continuously on X
to the right, IxG->I. Denote (x,g) h-> xg. We introduce the groupoid
structure on the setQ = X x G by defining the following structure maps. The

6

set of composable elements Q 2 = {((xg, h), (x, g) : x G X, g, h G G} C Q x Q ,
and the multiplication for ((xg,h), (x,g)) G Q 2 is given by

And also (x, g) 1 = (xg,g x ) ; d(x, g) = x, r(x, g) = xg, e(x) = (x, e^)- This
groupoid is called the transformation groupoid.

Let us recall (cf. [18]) the concept of right Haar System.

Definition 1 A right Haar system for the groupoid Q is a family {X x } x& x
of regular Borel measures defined on the sets Q x (which are locally compact
Hausdorff spaces ) such that the following three conditions are satisfied:

1. the support of each \ x is the set Q x ,

2. (continuity) for any f G C C (Q) the function f°, where

(xg,h) o ( x ,g) = (x,gh).

belongs to C C (X)

3. (right invariance) for any g G Q and any f G C C (Q),

f(ho g)d\ r{g )(h) = / f(u)dX d{g )(u)

7

One can also consider the family {X x } xeX of left-invariant measures, each \ x
being defined on the set Q x by the formula \ X (E) = X^E -1 ) for any Borel
subset E of Q x (where E^ 1 = {g G Q : g -1 G E}). Then the invariance
condition assumes the form:

Now, let [i be a regular Borel measure on X. We can consider the following
measures which will be called measures associated with /i: v — J X x d/i(x) on
Q, v~ x = J \ x dfi(x) and v 2 = J \ x x \ x dfi(x) on Q 2 .

If v = u^ 1 we say that the measure /i is a Q -invariant measure on X .

Definition 2 A topological groupoid Q on X is called locally trivial if there
exist a point x G X , an open cover {Ui} of X and continuous maps s x>i :
Ui — > Q x such that r o = i<i^ /or aZZ i.

Proposition 1 Assume that Q is a locally trivial groupoid on X and X is
second countable space. Let n be a regular Borel measure on X . Then

1. Q is transitive,

2. all isotropy groups of Q are isomorphic with each other,

3. for every y G X there exist an open cover {Vj} of X and continuous
maps s y j :Vj — >■ Q y such that r o Sj = id Vj ,

8

4- for every x G X there exists a section s x : X — > Q x which is fi-
measurable, i.e., for every Borel set B in Q x , s~ 1 (-B) is fi-measurable
subset of X ,

5. if the measure fi has the property that fi(A) = fi(A) for every ji-
measurable subset A of X, then the section s x is ji — a.e. continuous on
X.

Proof:

1. Let yi,|/2 £ X. Suppose that yi G U\ and y 2 G U 2 - Then r(si(yx)) = y\
and r(s 2 (y 2 )) = 1/2- But g = s^y^os^yi)- 1 has the property d{g) = y x
and r(g) = y 2 , and this means that Q is transitive.

2. For x,y G X let g yx be an element of Q such that d(g yx ) = x and
r i.9yx) = V- Then we have an isomorphism of the isotropy groups Q%
and Qv given by the formula ^97^3^070 g~l e Q v y .

3. Let g xy be an element of Q such that d(g xy ) — y and r(g xy ) — x. Then
in the fiber Q y we can simply define s y ^{z) = s Xji (z) o g x y for z G C/j.

4. Since X is second countable space, we can take a countable cover
{Ui}i =lj 2,... of the space X. Let us define s x (z) = Si(z) for z G U 1:

9

s x (z) = s 2 (z) for z G U 2 \ Ut, . . . , s x {z) = s n (z) for z e U n \(U 1 UU 2 U
. . . U n -i) etc.

It is easily seen that s x (z) is measurable.

5. The set of discontinuity of s x is contained in the union of sets dUi =
Ui \ Ui, i = 1, 2, which is of measure zero.

This ends the proof.

From now on we assume that considered groupoids are locally compact
and satisfy the assumptions of Proposition 1. It is known that in the case of
any locally trivial groupoid there exists a right Haar system (see [H]). Let
us choose a collection of sections {s x } x& x, defined by Proposition 1.3, and
denote by d^ y the right Haar measure on the isotropy group T y . (From the
assumption that the groupoid Q is locally compact it follows that all isotropy
groups are locally compact and have Haar measures.)

Definition 3 A right Haar system {\ x } xt =x on the groupoid Q is called con-
sistent with a Borel regular measure fi on the base space X if, for every
x G X and any f G C C {Q X ),

10

The above formula gives an explicit construction of the right Haar system
for many classes of groupoids (see, Section 3 for pair groupoid, and Section
6 for transformation groupoid).

Let us recall the concept of groupoid representation [IE],[2T]. It involves
a Hilbert bundle H over X, H = (X, {H x } xe x, //) (Dixmier [5] uses for it
the name of //-measurable field of Hilbert spaces over X). Here all Hilbert
spaces H x are assumed to be separable.

Let /i be a ^-invariant measure on X, and v and v 2 the associated mea-
sures on Q.

Definition 4 A unitary representation of a groupoid Q is the pair (14, H)
where H is a Hilbert bundle over X and U = {U(g)} ge g is a family of
unitary maps U(g) : H d ^ — > H r ^ such that:

1. U (e(x)) = idn x for all x E X ,

2.

U(g) o U(h) = U(g o h) for v 2 - a.e. (g, h) e Q 2

3.

U(g- 1 ) = U(g)- 1 for v -a.e. g eg

4

For every (ft, ip € L 2

(X,H,//)

G3g^(U(g)4>(d(g)),i;(r(g))) rig) eC

11

is v -measurable on Q. (Here L 2 (X, H,/i) denotes the space of square-
integrable sections of the bundle H 7 and (•, -) x denotes the scalar prod-
uct in the Hilbert space H x .)

3 Elementary properties of representations
of groupoids.

Definition 5 Unitary representations (Wi,Hi) and {U 2 , H 2 ) of a groupoid

Q are said to be unitarily equivalent if there exists a family {A x } x( zx of

isomorphisms of Hilbert spaces A x : Hi x — > H 2x ,x G X such that for every

x,y G X and for u—a.e. g G Q% the following diagram commutes

n\ x > rl\y

A X Ay

H2x > H 2 y

U 2 (g)

Definition 6 Let (U, H) be an unitary representation of Q and let Hi be a
Hilbert subbundle of H. We say that Hi is Q -invariant if U{g)H^ x C H^ y
for every x,y G X and for u—a.e. g G Q y x . Then the representation (W, Hi)
is called a subrepresentation of (U, H). A subrepresentation (U, Hi) is called
proper if Hi is proper subbundle of H 7 i.e. Hi ^ H and Hi is not null
space bundle.

12

Definition 7 A unitary representation (14, H) is called irreducible if it has

no proper subrepresentations.

Example 3 Let H = X x H be a trivial Hilbert bundle over X with fiber H .
For g G G%, x,y G X define an operator of the representation
U(g) : {x} x H — > {y} x H by U(g)(x,h) = (y,h). Such representation
(14, H) of the groupoid Q is called trivial representation. A trivial represen-
tation is irreducible if and only if it is one- dimensional, i.e., dimH = 1.

Example 4 Let H x = L 2 (Q x ,d\ x ), for x G X, be a Hilbert space of square
\ x -integrable functions on Q x , and for g G Q^.,x,y G X and f G H x define
U(g) :H x ^H y by

(U(g)f)(g 1 ) = f(g 1 og),

for g 1 G Q y .

A representation (14, H) is called regular representation of the groupoid

sua.

Now let us consider the regular representation of a pair groupoid Q = X x X.
Let fj, be a regular Borel measure on X. Let us define a right Haar system of
measures {[i x } x <zx on the pair groupoid Q , \i x being given on Q 0tX = X x {x}
by the formula n x (f) = f Qox f(y, x)dfi(y).

13

Then we have a simple invariance condition:

/ f[(y,x) o (x,z)]dfi{y) = / f(y,z)dfi{y).

For each x G X the Hilbert space L 2 (G 0x ,d[i x ) is obviously isomorphic
to L 2 (X).

Example 5 The regular representation of a pair groupoid Qq in the Hilbert
bundle {L 2 (^ ,i)}xex over X is given by the following family of operators
{U (g)}, 9 e Go, x , x,y G X

[U {g)f](z,y) = f(z,x)

where z G X .

Let us observe that the regular representation of pair groupoid is equivalent
to trivial representation in the trivial Hilbert bundle X x L 2 (X).

Now, we shall introduce the quotient groupoid Q /V (cf. [12]) and consider
its representations.

Let T be the isotropy groupoid of a groupoid Q. Let us define an equiva-
lence relation ~ on Q, for g,h G Q,

g ~ h <^ there exist 7i G T such that (7^ g) G Q 2 and 71 o g — h.

14

Let us notice that if 72 G Td( g ) then also g ~ g o 72 . Indeed, g o 72 =
72 S 1-1 9 = 7i 9 where 71 = # o 72 o g' 1 e r r(g) .

Denote the equivalence class of g e £ by [g], and the set of such equiv-
alence classes by £?/r. Then we can introduce the groupoid structure on
Q/T. The structure maps d and f, the multiplication, the inverse and the
identity section e are given by d[g\ = d(g), r([g\) = r(g), [g] o [h] = [g o h],
for (g,h) e Q 2 , [gY 1 = [g' 1 ], i(x) = [e(x)], respectively.

It easy to see that the canonical projection p Q/T is a homomor-

phism of (topological) groupoids (in Q/T we choose the quotient topology).
Let us denote by Qo the pair groupoid Q = X x X over the base X. Recall
that in Q we have d (x, y) = y, r (x, y) = x, (x, y) o (y, z) = (x, z), (x, y)' 1 =
(y,x) and e (x) = (x,x).

We observe that the quotient groupoid Q/T coincides with Q .

Proposition 2 The map $ : Q/T — > Q , given by $([#]) = (r(g),d(g)), is
an isomorphism of groupoids over X .

Proof: It is clear that, for g, h e Q 2 , one has $([<?] o [h]) = (r(g), d(h)) =
(r(g),d(g)) o (r(h),d(h)) = $([<?]) o $([/*]). Also ^([g]' 1 ) = (d(g),r(g)) =
($([(?])) _1 . Thus $ is a groupoid homomorphism. It is clear that <£> maps
Q/T onto Qq. Moreover, if $([(?]) is a unit element in Q , i.e., $([<?]) = (x,x)

15

for an element x G X, then d(g) — r(g) = x, i.e., g G T and [g] in a unit in
(//T. This means that <£> is an isomorphism, o

Now, let us assume that a representation (U, H) of the groupoid Q is
T— invariant, i.e., (7(705) = £/(#) for every g <E G, 7 G T, (7,5) G
Then it is easily seen that one has a unitary representation (Wo, H) =
({U ([g})}[ g ](zg/r, H) of the groupoid £/T formed by the family of operators

U ([g}) = U(g) :H x ^H y

for every g such that d(g) = x, r(g) = y.

Example 6 Let H x = L^{Q X ) be a Hilbert space of Y— invariant and
ji— square integrable functions on Q x , i.e. such that, for Borel-measurable
functions f on Q x , f(jog) = f(g) for every g G Q x , 7 G Y, (7,5) G Q 2 , and
fx \f(.9)\ 2 dfi(r(g)) < 00. It is clear that the space Ll(Q x ) is isomorphic to
L 2 (X). Define for every g G G%, x,y G X an operator U(g) : H x — >■ H y by

(U(g)f)(g 1 ) = f(g l og)

for f G H x , g 1 G Q y . In such a manner we obtain a Y -invariant unitary
representation of the groupoid Q which is called a quasi-regular representa-
tion. Let us observe that the corresponding representation U$ of the quotient

16

groupoid Q/T coincides with the regular representation of the pair groupoid

Go-

4 Induced representations of the groupoid Q.

In this section, we define the representation of Q induced by a representation
of the isotropy subgroupoid T. From now on we assume that on the groupoid
Q there exists a right Haar system {\ x } xe x consistent with Borel regular
measure // on X.

First, we have to construct an appropriate Hilbert bundle.

Assume that there is given a unitary representation (r, W) of the sub-
groupoid T. Here W is a Hilbert bundle over X. Let W x denote a fiber
over x G X which is a Hilbert space with the scalar product (•, -) x , and let
W = U xe xW x denote the total space of the bundle W.

Let us define, for every x G X, the space W x of W- valued functions F
defined on the set Q x satisfying the following four conditions:

1. F(g) G W r (g) for every g G Q x ,

2. for every //-Borel measurable r-section s x : X — > Q x (see Proposition
1) the composition F o s x is a //-measurable section of the bundle W,

17

3. F( 7 o g) = r^)F(g) for g G Q x , 7 G r r(ff) ,

4. /<F(s a; ( J /)),F(s x (j/))) 1 ,dMj/)<oo.

If we identify two functions F, F' G W x satisfying

I <(F - F')My)), (F - F^My)))^) = 0,

we can introduce the scalar product (•, -) x in the space W x

(F 1 ,F 2 ) X = J (F 1 (s x (y)),F 2 (s x (y))) y d^y)

where s x is the section determined by Proposition 1, part 3.

The spaces W x , x G X, with these scalar products are Hilbert spaces. It is
easily seen that they are isomorphic to the Hilbert space L 2 (X, W) of square-
integrables sections of the bundle W. Now, let us denote W = {W x } xe x-
It is a Hilbert bundle over X. We define a unitary representation of the
groupoid Q in the Hilbert bundle W in the following way

Definition 8 The representation of the groupoid Q induced by the represen-
tation (r, W) of the subgroupoid T is the pair {U T , W) where, for g G Q y x , we
define U T (g) : W x -)• W y by

(U T (g )F)(g) = F(gog ).
18

It is clear that (U T , W) is a unitary groupoid representation.
Sometimes we shall use the notation U T = Ind^(r).

5 Systems of imprimitivity.

For a given Hilbert space W we can consider the Hilbert space L 2 (X, W ) of
square integrable W - valued functions on X. In the space L 2 (X,W ) one
has a representation of the commutative algebra L°°(X) given by the mul-
tiplication operators by the function: L°°(X) 3 f ->■ 7T (/) G B(L 2 (X,W ))
where, for z G X,

We shall call 7r £/ie natural representation of L°°(X) in L 2 (X, W ).

Now, let us consider a representation U of the groupoid Q in & Hilbert
bundle H over X. We assume that for /i - a.e. x G X there exists a Hilbert
space W x with a scalar product (•, -) x such that the spaces W x are isomorphic
with each other. Let us assume that, for /j, - a.e. x, the fiber H x of the bun-
dle H is isomorphic to L 2 (X, W x ). We shall simply write H x = L 2 (X, W x ),
and U(g) : L 2 (X,W X ) ->■ L 2 (A",W W ) for 5 G ^. It is clear that the collec-
tion of the spaces {L 2 (X,W x )} x( zx forms a Hilbert bundle over X which is

19

isomorphic to the bundle H.

Definition 9 We say that there exists a system of imprimitivity (U, n) for

the representation (U, H) of the groupoid Q if

1. the representation (U, H) satisfies the above assumption (H x =
L 2 (X,W X ) for fi -a.e. xeX),

2. 7r = {j x ) x& x is the family of natural representations of the algebra
L°°(X) in the Hilbert spaces L 2 (X, W x ),

3. for every f G L°°(X), and for /i - a.e. x,y G X, and v - a.e. g G Q%

U{g)-n x {f)U{g~ 1 ) = ir y (f).

Example 7 Let (U, H) be the quasi-regular representation of the groupoid
Q , defined in Example 6. Then there exists a system of imprimitivity (U, ir)
forU. Indeed, for f G L°°(X), tp G H y , g G Q%, h G Q y , we have

(U(g)7r x (f)U(g- 1 )^(h) = (nMUig-^hog) =

= f(r(h o g^Uig" 1 )^ o g) = f{r{h))^{h) = ir v (f)^(h).

The quasi-regular representation can be understood as induced by a trivial
one- dimensional representation of the subgroupoid T.

20

We are now in a position to state our main theorem (the imprimitivity
theorem for groupoids):

Theorem 1 If, for a representation (U, H), there exists a system of imprim-
itivity (II, 7r) then the representation U is equivalent to the representation U T
induced by some representation (r, W) of the subgroupoid T.

Let us observe that, for 7 e T x — Q x , condition 3 of the Definition 9 of
the imprimitivity system reduces to the following one

U{T)* x {f)U{T x ) = **(/)■
Let us denote by Mo :X the following subalgebra in B(H X ) = B(L 2 (X, W x ))\

M , x = {c/( 7 ) : 7 e rj.

Then we have, for /x - a.e. x G X, 7r :c (L 00 (X)) C A^o x , where M' 0x
denotes the commutant of the algebra Mo, x in B(H X ).

Definition 10 The system of imprimitivity (U, n) is irreducible if, for \i -
a.e. xeX, 7r x (L°°(X)) = M' 0x .

Theorem 2 If for a representation (U, H) there exists an irreducible sys-
tem of imprimitivity (U, n) then the representation U is equivalent to the
representation U T induced by some irreducible representation (r, W) of the
subgroupoid Y .

21

First, let us notice that from the fact that all operators U(j), 7 G T x ,
commute with all ir x (f), f G L°°(X), it follows that U(^y) are decomposable
(see [5], part II, 2.5). This means that, for fi — a.e. y G X, there exists
an operator U(^) y G B(W X ) such that, for ip G L 2 (X,W X ), (U(j)ip)(y) =
U( r y) y (i/;(y)). It is easily seen that all U(j) y are unitary (cf. |5], II. 2, Ex.2).
We can prove even more.

Lemma 1 If for a representation (U, H) there exists a system of imprimi-
tivity, then

1. there exists a unitary representation (t x , W x ) of the group T x such that
T x(l) = U(j) for every 7 e T x and fi - a.e. x G X. (In particular it
means that the function X 3 y — > U(j) y G B(H X ) is a constant field of
operators ),

2. we can define a representation (r, W) of the subgroupoid T such that,
for 7 G T x , r(i) = 7-3.(7),

3. if the system of imprimitivity (U, it) is irreducible then (r, W) is an
irreducible representation of V , i.e., for fi - a.e. x G X , the represen-
tations (t x ,W x ) of the groups T x are irreducible.

22

Proof: Notice that the Hilbert space L 2 (X,W X ) is isomorphic to the
tensor product of Hilbert spaces L 2 (X) (£)W X . A decomposable operator in
such a space has the form [A(ip ® h)](y) = ip(y) <S> A y h. We have to show
that it is of the form id L 2 <S> A , where A G B{W X ). Let {/0i}t=i,2,... De
an orthonormal basis of the space L 2 (X). Consider the unitary operators
Uij in the space L 2 (X) defined by Uijipj — If A is decomposable then
A commutes with all operators of the form Uij <g> idjj x - Then it follows that
A = iduz®AQ by Lemma 2 ([5], section 1.2.3). For the operators U(y), , y G T x
let us denote by t x (j) the operators W x — >■ W x such that U(j) = id L 2 ®T x (y).
It is clear that all t x (j) are unitary in W x , and t x (ji o 72) = t x (ji) o ^(72)
for 71,72 G T. Thus r x is a unitary representation of the group T x in the
Hilbert space W x . This ends the proof of part 1.

Now the assertion 2 of the Lemma is obvious.

To obtain part 3 it is sufficient to see that the condition of irreducibility of
the imprimitivity system implies that only operators of the form Xidw x , (A G
C) commute with all 7^(7), 7 G T x . But by Schur's lemma it follows that the
representation t x of T x is irreducible, o

The next lemma gives us more properties of the representation (r, W) of
the groupoid T as well as of the representation (U, H) that has a system of

23

imprimitivity.
Lemma 2

1. The representations t x ,x G X are equivalent to each other, as repre-
sentations of isomorphic groups T x .

2. The operators U(g) : H x ->■ H y , where H x = L 2 (X,W X ), H y =
L 2 (Y, W y ) for g G Q y x , are decomposable, i.e., there exist unitary op-
erators U°(g) :W X ^W y such that for ip G L 2 (X, W x ) and, for z G X ,

(U(g)^)(z) = (U°(g))Mz)).
Moreover, the operator U°(g) : W x — > W y does not depend of z G X .

Proof: First we shall prove part 2. First of all, let us notice that all
spaces W x , for /i - a.e. x G X, are isomorphic to each other as Hilbert spaces.
Denote by i y x : W x — » W y the isomorphism and define the unitary map R y x :
L 2 (X,W X ) -> L 2 (X,W y ) by (RW)(z) = i%(^(z)), i> G L 2 (X,W x ),z G X.
Consider the composition of unitary maps U(g) o (R^Y 1 : L 2 (X,W y ) — >■
L 2 (X, Wj,) where (7 G By using the property of the imprimitivity system
for U(g), we obtain

U(g) o (it*)" 1 o = n y (f) o [/(«?) o (Rl)^

24

for / G L°°(X).

This means that the operator ?7((yf)o(i?^) _1 is decomposable in L 2 (X, W y ).
But (R y x ) is a decomposable map by definition, therefore U (g) is decompos-
able as the composition of decomposable maps. As in the proof of Lemma 1
we conclude that U°(g) does not depend of z G X and is unitary.

To prove part 1 let us first observe that the isotropy groups T^, are
isomorphic to each other x G X. Indeed, taking an element g G Q y x we define
the isomorphism i : T x — > T y by the formula 2(7) = 5070 g^ 1 for 7 G T x .
Now, we have U(i(^)) = %d L i <g> 7^(2(7)) as in the proof of Lemma 1. On the
other hand, £7(2(7)) = U(g) o ^(7) o L% _1 ) = (id L 2 <g> £/°(#)) o (id L 2 0^(7)) o
(irf L2 <g> U°(g)- 1 ) = id L 2 <g> (C/°(#) o 7-3(7) o f/ ^)" 1 ). Therefore, we have
T y {i{^)) = U°(g) o 7-21(7) o i7 (<7) -1 , but this means that the representations
T y and t x are equivalent.

Now, we are in a position to give proofs of Theorems 1 and 2.

Proof. Let us consider the spaces {W x } x ex, introduced in Section 1,
connected to the representation r of Lemma 1 and the corresponding induced
representation U T . We shall show that the representation (U, H) is equivalent
to (U T , W). We define a family of isomorphisms of Hilbert spaces J x : H x — >■

25

W x for fj, - a.e. x G X. Since H x = L 2 (X,W X ), for ip G if^, g G (? x , and
r {g) — £/> we P u t F{g) = (J x i/))(g) = (U(g)(i/)))(y). The definition is correct
since by Lemma 2 we have (U(g)ip)(y) = U°(g)(ip(y)), and U°(g) does not
depend of y G X. Since C/(#)V> G L\X,W y ), therefore [U(g)(ijj)](y) G W„.
Also it is clear that F(j o g) — r(j)(F(g)) for 7 G 1^. To see the square-
integrability let us write

J(F(s x (y)),FMy))) y dfi(y) =

= j(u\s x {y))my),u\ Sx {y))my))yd^y) = f&(y)^(y))vMy) =

=11 i> \\h x < 00.

This also shows that J x are unitary maps and are injective. To see
that J x map onto W x , we can give the formula for J" 1 : (J x 1 F)(y) =
(U°(g))~ 1 (F(g)) where F G W x and g G Q y x . Then the right-hand side does
not change if we take other element g\ G Q y x . Indeed, since g 1 = 'yog,
for an element 7 G T y , therefore we have (£7° (7 o g))~ 1 (F(y o g)) =
((f/ (( ? ))- 1 (r(7))- 1 (r(7))(F(^)) = (U° (g))' 1 (F (g)) . This shows that J x ,
x G X, are isomorphisms of Hilbert spaces. Now we can see that J x are
intertwining maps for the representations U and U T , i.e., that the following

26

diagram commutes

rr Vi9) rr

n x > n z

U T (g)

for /i-a.e. x,z G X and 1/ - a.e. g G Let ip G i^. Then, for h G
0jjf, we have [(J z U(<,))ty)](h) = [{U{h){U{g))^)]{y) = U{hog)^(y)) =
U°(h o g)(iP(y)). On the other hand, U T (g)J x (^)(h) = [J x (ip)](h o =
[C/(/i o g r )(V')] (?/)■ This ends the proof of Theorem 1.

The theorem 2 is now a simple consequence of Theorem 1 and Lemma 1,

part 3.

6 Representations of the transformation
groupoid Q = X x G, X = K\G

As an introduction to this section we recall the concept of induced represen-
tation in Mackey sense (cf [15], [13], [22]) of a Lie group G by a unitary
representation (L, V) of its closed subgroup K defined in a Hilbert space V.

We assume, for simplicity, that X = K\G has a G-invariant measure /i.
We consider Hl, a Hilbert space consisting of measurable functions on G

27

with values in V, such that

( f ) (hg) = L(h)<P(g),heK,

and

< oo

where [g] denotes the image of g in X under the projection G — > K\G = X.
We introduce the inner product

Then we define the representation U L of G on T-Ll given by the formula

It is easily seen that U L is unitary. The representation (U l ,V,l) is called
induced by the representation L of K.

Let G be a noncompact Lie group and K its compact subgroup. We
assume that G is unimodular. Then the homogeneous space X = K\G is a
G— manifold with right action of the group G: X x G 3 (x, g) (->■ xg e X.

As above, we assume that there exists a G-invariant measure fi on the
space X, i.e., for / e C C (G) we have

(t/ L G?)/)G?o) = /(aw), weGJe

28

We shall consider the structure of transformation groupoid on Q = X x G
(cf. Example 2) and construct a right Haar system on Q consistent with the
measure \i.

Let us denote

G x = {(x,g)eg:geG},

G y = {(yg-\g)eg:geG}.

Let us also denote the isotropy group Q% by T x , T x = {(x,k) : k G K x },
where K x is a subgroup of G of the form K x = g^Kgo where go G G is
an element of the coset x (x — [go])- Indeed, for k x e K x we have xk x =
[golg^kgo = [kg ] = x.

Lemma 3 Let s be a Borel section of the principal bundle G — > K\G = X ,
i. e., \sq{x)\ = Ks (x) = x. Then

1. for every x G X there exists a section s x : X — >■ g x with respect to the
map r, i.e., r(s x (y)) = y,

2. every element g G Q x can be represented as g = k o s x (y) where k G

r = G y

Proof:

29

1. Let o G X denotes the origin point, i.e., o = [k], k G K. Then we have
[fcso(a;)] = x or, equivalently, osq(x) = x. Analogously, oso(y) = y for
y G X. Thus we can define the section s x : X — > Q x by the formula
s x (y) = (x, s (x) _1 s (y))- It is clear that r(s x (y)) = y.

2. Let us observe that the product ko s x (y), where k e T y = Q v y , is of the
form

(y, soiy^ksoiy)) o (x, so(x)' 1 s (y)) = (x, s (x) _1 A;so(y))-

We have to show that if g = (x, g) with g such that xg = y, then
there exists k G K such that g = s ( ;r )" 1 ^ s o(y) or > equivalently,
so^gsoiyy 1 = k. Now, oso^gsoiyy 1 = xgs Q (y)~ 1 = ys Q {y)' 1 = o.
But the isotropy group of the origin point o is equal to K, what means
so^gsoiyy 1 G K. o

Now, for a function / G C C (Q X ), let us define f x (y,k) = f(x,s (x)~ 1 ks (y)),
and

/ /(g)dA jB (g)= / / f x (y,k)dkdiM(y).
Jg x Jx Jk

Proposition 3 The collection {\ x } x <=x is a right Haar system on the
groupoid Q consistent with the measure ji on X .

30

Proof: We have to show the right-invariance of the system {A^}. Let
z = r (go)j go = (x,go) such that xg = z and let us compute

/ f(s°So)d\ z (g) = / f[(z,s (z)~ 1 ks (y))o(x,g )]dkdn(y) =
Jg z Jx Jk

j j f(x,gos (z)~ 1 kso(y))dkdfji(y).

Let us observe that oso^goSo^z)' 1 = zs Q (z)~ l = o, thus
s o( a; )fl , o s o(- 2 ) _1 ^ K an d s o( x )9o s o( z ) — ki e K. But this means that
9oSo{z)~ 1 k = so(x)~ 1 ki. Thus, continuing the computation, we have

f(x,g s (z) ks (y))dkdfi(y) = f(x,s (x) kiSo(y))dkidn{y)

f(g)d\ x (g). o

Now, we shall consider representations of the isotropy subgroupoid T. As
we have seen, T = \J x ex{ x } x K x with K x = g~ l Kg and g e G such that its
coset in X is equal to x ([<?] = rr). We can use g = so(x).

Let (r, W) be a unitary representation of the groupoid T in a Hilbert
bundle W = {W x } xeX .

Definition 11 A representation (r, W) is called X- consistent if there exist
a unitary representation (r , Wo) of the group K and a family of Hilbert space

31

isomorphisms

A x : Wo -> W x , x e X
such that, for 7 <ET X of the form 7 = (x, so(x)~ 1 kso(x)),

r( 7 ) = A x r {k)A-\

Proposition 4 Let (U, W) fre a unitary representation of the groupoid Q.
Then the restriction (r, W) to £/ie subgroupoid T of the representation
(U,W), given by the formula t{j) = U{j) for 7 6 T, is a X-consistent
representation ofY.

Proof. We can write 7 = (x, s (a:) _1 A;so(a;)) = (o, s (x)) o (o, A;) o
(o^soix))- 1 . Then C/( 7 ) = U((o, s (x)))U((o, k))U((o, s (x)))-\ Let us
denote Ar = C/((o, s (x))) and r (A;) = U((o,k)), Wo = W Q . Then it is clear
that {A^j^gx an d r satisfies the conditions of X-consistent representation.

o

In the sequel we shall consider the representation of the groupoid Q =
X x G induced by X - consistent representation (r, W) of the subgroupoid
T, and we shall establish its connection with the induced representation in
the Mackey sense of the group G. We use the notation of section 3. Now

32

condition 3 of the definition of the space W x assumes the form

F(l°(x,g)) = T{j)F{x,g)

where x,y G X, y = xg, g G G, 7 G T y = {y} x K y . Thus we have
7 = (y, sod/) _1 /cso(y)) for an element k G K . Then, by the definition of
X-consistent representation, we can write

-^(7 ( x , g)) — (A y T (k)A y 1 )F(x, g).

Let introduce a function : G — >■ W defined by the formula <f>(ks (y)) =
A~ 1 (F(a;, SQ^x^ksQ^y))). Then the function <p has the property <f>(kg) =
ro(k)<f>(g).

It is sufficient to check the above formula for g — s (y). If 7 =
(y, s d/) _1 ^so(y)) then we have <f>(kg) = A" 1 ^ (7 {x, so(x)- 1 s (y)))) =
A-^AyTo^A-^Fix^soiy)) = T (k)(f)(g).

We shall use the notation (L, W ) for the unitary representation of the
group K in the space Wq, L = r . Thus we have 4>{kg) = L(k)<f>(g) and we can
consider the Hilbert space Hl introduced above as well as the representation
(?7 l ,'Hl) of the group G induced in the sense of Mackey by L from the
subgroup K.

The following theorem establishes a connection of the induced represen-

33

tation (U T ,W) of the groupoid Q with the representation (U L ,1-L L ) of the
group G.

Denote by R g , g G G, the following operator acting in the space W X: x G
X,y = xg,

(R g F)(x,h) = (A xh A^ g )(F(x,hg)).

Then we have the family of unitary G-representations (R, W x ),x G X.
(The unitarity follows from the fact that the measure \i is G-invariant and
the operators A x h,A x h g are Hilbert space isomorphisms.)

Theorem 3

1. For every x G X the G -representation (R,W X ) is unitarily equivalent
to the induced representation (U 1 ^^.

2. All representations (R,W X ), x G X , are unitarily equivalent to each
other. The equivalence is given by the operators 11 : W x —> W y ,

{IlF){y,s,{y)- l ks Q (z)) = (A y A~ 1 )(F(x, so(x)~ 1 kso(z))),

x,y G X.

Proof.

34

1. We define the linear map J x : W x — > Hl by (J x F)(g) = <f>(ks (y)) =
Ay 1 (F(x, so(x)~ 1 kso(y))) where g = kso(y). J x is a linear isomorphism
since A y is an isomorphism and it is easily seen that J x preserves scalar
products of W x and % L and so it is a Hilbert space isomorphism. To
see that it defines an equivalence of representations, we have to show
that, for g G G, the following diagram is commutative

Jx

Jx

U L (g)

Let us compute (U L (g)J x )(F)(h). It is sufficient to take h = s (y) and
to notice that each g G G can be written in the form g = So(y)~ 1 ks (z),
for z G X, z = yg and an element k G K.

(U^Soiy^ksoiz^J^F^Soiy)) = (J x F)(ks (z)) =

= L{k)A-\F{x,s Q {x)- l s {z))).

On the other hand

(J x R g )(F)(s (y)) = A y l ((R g F)(x,s (xy l s (y))) =

= A^ 1 (A y A; 1 )(F(x,s (x)- 1 ks (z))) = A; 1 r(7)(F(x, So (a;)- 1 S oW)) =
= A~ 1 A z T (k)A~ 1 (F(x, sq(x)~ 1 sq(z))) = L{k)A-\F{x, s^x)' 1 s (z))).

35

2. Now it is a simple observation that 1% = J y 1 J X . o

7 A physical picture. A concept of particle
in the representation theory framework

In papers ([7], [9], [8], [E]) we have studied a model unifying general
relativity and quantum mechanics based on noncommutative geometry The
principal structure of the model is provided by a transformation groupoid
G = E x G where G is the Lorentz group and E is the principal G-bundle
over the spacetime M (the total space of he bundle is formed by all Lorentz
frames at all points of M). We have defined a right action of G on E, and
the multiplication of elements of the groupoid is introduced as follows

(P9,9x) ° (P,g) = (P,99i),

V e E,g,gi G G.

The model is reduced to the usual quantum mechanics when an act of
measurement is performed. Then we choose a frame p G E which represents a
reference frame in which the measurement is done. In the sequel we consider
the situation when we want to observe a particle from a different reference

36

frame situated at a fixed point x G M. In such a case, our groupoid reduces
to the groupoid G = E x x G where E x is the fiber of the bundle E over x. The
groupoid G is transitive and its isotropy subgroupoid is trivial T = E x x {e},
where e G G is the neutral element of the group G.

A representation (r, W) of T is realized in a trivial Hilbert bundle W =
{W p } p€ E x with W p = C and t(j) = id\y v for 7 = (p, e). Therefore, the
induced representation ([/ r , VF) of G is simply a regular representation (cf
...)• Indeed, for p G E, we have G p = {(p,g),g G G}, W p = L 2 (G P ) ^ L 2 (G)
and, for F G W p , h e G,

(U T (p,g)F)(pg,h) = F((pg,h) o (p, g)) = F(p,gh).

Let us notice that this regular representation can serve to define random
operators on the groupoid and then to define the von Neumann algebra of
the groupoid (cf. [8], [19], [9]).

Now we pass to quantum mechanical momentum representation of a par-
ticle with the mass m. Having fixed (by an act of measurement) p G E, we
have reduced our initial space to {p} x G = G. But we want to consider
the energy-momentum space H of the particle, H = {(po,Pi,P2,P3) G R 4 :
Po ~ Pi ~ v\ ~ v\ — m }- We have an action of the group G = SL 2 (C) on the

37

hyperboloid H (see

To describe the action we identify H with the set H of hermitian 2x2-

matrices with determinant equal to m,

{P0,Pl,P2,P3) H>

V

po -P3 pi- m

P2 + iPl PO + P3

and we let to act g G G on H to the right in the following way,
~H 3 A h-> g*Ag e H. (It is clear that det(g*Ag) = detA = m). Next,
we see that the isotropy group of the element (po, 0,0,0), po = y/m is
equal to K = SU(2). Thus we deduce that the homogoneus space K\G is
diffeomorphic to H . We can take the phase space of a particle of the mass
m as the space Q = K\G x G = H x G and consider the algebraic structure
of transformation groupoid on it.

Let (U, W) be a unitary representation of the groupoid Q in a Hilbert
bundle W. Assume that there exists an imprimitivity system (U, tc) for
(U, W). We say that a particle of mass m is represented by the pair (U, tt).
We say that it is an elementary particle if the imprimitivity system (U, 7r)
is irreducible p2], [H]. Equivalently (on the strength of Theorem 1), we
can say that the particle is an induced representation (U T , W) where r is a

38

unitary representation of the isotropy subgroupoid T. In the same way, we
can say that the particle is elementary if the inducing representation r is
irreducible and, in turn, this means that the representation (L, Wo), L = t ,
of the group K = SU{2) is irreducible. Then the representation (L, Wq) is
called the spin of the particle.

References

[1] R. Bos, Continuous representations of groupoids, arXiv:math/0612639v3
[math.RT].

[2] R. Brown, From groups to groupoids, Bull. London Math. Soc. 19 (1987),
113-134.

[3] M.R. Buneci, Groupoid C*-algebras, Surveys in Mathematics and its
Applications, ISSN 1842-6298, 1 (2006), 71-98.

[4] A. Cannas da Silva and A. Weinstein, Geometric Models for Noncom-
mutative Algebras, American Mathematical Society, Berkeley, (1999).

[5] J. Dixmier, Von Neumann Algebras, North Holland Publ. Comp., Am-
sterdam, (1981).

39

[6] M. Heller, Z. Odrzygozdz , L. Pysiak, W. Sasin, Structure of Malicious
Singularities, Int. J. Theor. Phys., 42, (2003), 427 - 441.

[7] M. Heller, L. Pysiak, W. Sasin, Noncommutative unification of general
relativity and quantum mechanics, J. Math. Phys., 46 (2005), 122501-15.

[8] M. Heller, L. Pysiak, W. Sasin, Noncommutative dynamics of random
operators, Int. J. Theor. Phys., 44 (2005), 619-628.

[9] M. Heller, L. Pysiak, W. Sasin, Conceptual unification of gravity and
quanta, Int. J. Theor. Phys., 46 (2007), 2494- 2512.

[10] M. Heller, Z. Odrzygozdz , L. Pysiak, W. Sasin, Gravitational Aharonov-
Bohm Effect, Int. J. Theor. Phys., 47, (2008), 2566-2575.

[11] N.P. Landsman, Mathematical Topics between Classical and Quantum
Mechanics, Springer, New York, (1998).

[12] K.C.H. Mackenzie, Lie groupoids and Lie algebroids in Differential Ge-
ometry, London Math. Society Lecture Notes Series, 124, Cambridge
University Press, Cambridge, (1987).

[13] G.W.Mackey, The relationship between classical mechanics and quantum
mechanics, Contemporary Math., 214 (1998), 91-109.

40

[14] G.W.Mackey, Unitary group representations in physics, probability and
number theory, Benjamin, Reading, Mass., (1978).

[15] G.W.Mackey, Induced representations of locally compact groups I, II ,
Acta Math., 55 (1952), 101-139; 58 (1953), 193-221.

[16] G.W.Mackey, Imprimitivity for representations of locally compact
groups, Proc. Nat. Acad. Sci. U.S.A. 35 (1949), 537-545.

[17] J. A. Packer, Applications of the work of Stone and von Neumann to the
theory of wavelets, Contemporary Math., 365, (2004), pp. 253-279.

[18] A.L.T. Paterson, Groupoids, Inverse Semigroups, and Their Operators
Algebras, Birkhauser, Boston, (1999).

[19] L. Pysiak, Time Flow in a Noncommutative Regime, Internat. J. Theo-
ret. Phys., 46 (1), (2007), 16 - 30

[20] L. Pysiak, Groupoids, their representations and imprimitivity systems,
Demonstratio Mathematica 37 (2004), 661-670.

[21] J.N. Renault, A groupoid approach to C*-algebras, Lecture Notes in
Math. 793, Springer- Verlag, New York, (1980).

41

[22] M.E. Taylor, Noncommutative Harmonic Analysis, A. M.S., Providence,
(1986).

[23] A. Weinstein, Groupoids: unifying internal and external geometry, Con-
temporary Math. 282, (2001), 1-19.

[24] J. Westman, Harmonic analysis on groupoids, Pacific J. Math., 27
(1968), 621-632.

42